摘要
提出了一种基于视觉知识加工模型的目标识别方法.该加工模型结合目标定位、模板筛选和MFF-HMAX(Hierarchical model and X based on multi-feature fusion)方法对图像进行学习,形成相应的视觉知识库,并用于指导目标的识别.首先,利用Itti模型获取图像的显著区,结合视觉通路中What和Where通道的位置、大小等特征以及视觉知识库中的定位知识确定初期候选目标区域;然后,采用二步去噪处理获取候选目标区域,利用MFF-HMAX模型提取目标区域的颜色、亮度、纹理、轮廓、大小等知识特征,并采用特征融合思想将各项特征融合供目标识别;最后,与单一特征以及目前的流行方法进行对比实验,结果表明本文方法不仅具备较高的识别效果,同时能够模仿人脑学习视觉知识的过程形成视觉知识库.
A novel object recognition method based on visual knowledge processing model is presented. Combined with object localization, template screening and hierarchical model and X based on multi-feature fusion（MFF-HMAX） method,the visual knowledge processing model yields a visual knowledge base which can be used as a guide in object recognition.Firstly, significant areas of the image can be obtained via Itti model; according to these areas and ＂what＂ and ＂where＂information, such as location, size, etc., the candidate ob jects are conformed. Secondly, MFF-HMAX model is used to extract various features, like color, intensity, texture, contour, size, etc., from the objects denoised by the two-step denoising process. After multi-feature fusion, the features can be used in object recognition. Finally, the method is tested and compared with single feature method and current popular methods. The results show that this method can not only get good performance in improving accuracy of object detection, but also yield a base of visual knowledge by imitating the forming process in human brain.
出处
《自动化学报》
EI
CSCD
北大核心
2016年第5期760-770,共11页
Acta Automatica Sinica
基金
国家自然科学基金(31170952)
国家海洋局项目(201305026)
上海海事大学优秀博士学位论文培育项目(2014bxlp005)
上海海事大学研究生创新基金项目(2014ycx047)资助~~
关键词
目标识别
视觉知识
自顶向下机制
自底向上机制
Object recognition
visual knowledge
top-down（T-D） mechanism
bottom-up（B-U） mechanism
作者简介
随婷婷 上海海事大学博十研究生．2013年获得上海海事大学信息学院硕士学位．主要研究方向为视觉榆测，视觉注意力模型，人工智能，数据挖掘．本文通信作者．E-mail：suisui61@163．com
王晓峰 博士，上海海事人学教授．主要研究方向为人工智能，数据挖掘与知识发现．E-mail：xfwang@shmtu．edu．cn
